Stable Diffusion Modelscope Basics - Make Cool Animations!!!

[@yonatandor7389]

Thank you for the information!

[@ysy69]

Thanks for making this video. Does Modelscope allow you to use your own custom ckpt ?

[@neoProfessor]

Thanks.
Good question, I think its probably best to ask on the discussion section on the GitHub page
github.com/deforum-art/sd-webui-text2video/discussions
I did also find this training script, but I personally haven't used it
huggingface.co/damo-vilab/modelscope-damo-text-to-video-synthesis/discussions/11
Hope it helps

[@justinwking]

look for the animov model scope model, after fine tuning, the shutter stock logo disapears

[@liquidmind]

SO ITS ANOTHER MODEL JUST FOR THAT EXTENSION...?

[@pjonesdotca]

There is a balance between the video dimensions and the fps. You can go with 256x192 and get a few more frames out (say 32 or 48) with a smaller GPU. At 8fps I was able to generate almost ten seconds of footage. The caveat here is that it seems that the temporal consistency doesn't last more than 3s so it will get super weird after 3s

[@neoProfessor]

Interesting, will try this out. Perhaps you could use some form of frame interpolation in an external program to smooth out the low fps.

[@pjonesdotca]

@@neoProfessor that's the goal. Push comes to shove I'll install FrameFlow but, I suspect there is already a A1111 extension available for that.

[@neoProfessor]

Just realized I forget to set the ETA setting back to 0 after setting it to 0.7. Explains why the comparison video after look slightly different from one another.

[@realmakebelieve]

I got this error anyone know what I can do? AttributeError: can't delete attribute 'sd_model'
Exception occurred: can't delete attribute 'sd_model'

[@jeremygalimberti4792]

pretty sure its too many characters/ words. I get it when im using too many loras or trying longer prompts, cant seem to get more than 50 words but sometimes less

[@onioncuttingninja]

how far we have come